2021/04/26 by Ashkan Kazemi, Kazemi, Ashkan, Zehua Li +5 · 1 citation
Social Sciences · Computer Science · #Misinformation and Its Impacts #Topic Modeling #Advanced Text Analysis Techniques
paper · pdf · doi:10.48550/arxiv.2104.12918
In this paper, we explore the construction of natural language explanations\nfor news claims, with the goal of assisting fact-checking and news evaluation\napplications. We experiment with two methods: (1) an extractive method based on\nBiased TextRank -- a resource-effective unsupervised graph-based algorithm for\ncontent extraction; and (2) an abstractive method based on the GPT-2 language\nmodel. We perform comparative evaluations on two misinformation datasets in the\npolitical and health news domains, and find that the extractive method shows\nthe most promise.\n